Reward tampering problems and solutions in reinforcement learning: a causal influence diagram perspective.

Can humans get arbitrarily capable reinforcement learning (RL) agents to do their bidding? Or will sufficiently capable RL agents always find ways to bypass their intended objectives by shortcutting their reward signal? This question impacts how far RL can be scaled, and whether alternative paradigm...

Descripción completa

Detalles Bibliográficos

Publicado en:	Synthese Vol. 198; no. 27; pp. 6435 - 6468
Autores principales:	Everitt, Tom, Hutter, Marcus, Kumar, Ramana, Krakovna, Victoria
Formato:	Artículo
Publicado:	Springer Nature Nov2021 Supplement 27
Materias:	Reinforcement learning Artificial intelligence Reward (Psychology) Decision theory
Acceso en línea:	Ver este registro en EBSCOhost